Anthropic’s Claude model has achieved a groundbreaking milestone by successfully solving the highly complex nine-loop scattering amplitude calculation problem in high-energy physics. Remarkably, this feat was accomplished with minimal human intervention, at a cost of just a few thousand dollars, and using straightforward prompts. The model computed the nine-loop result for the six-particle amplitude in planar N=4 super Yang-Mills theory, surpassing the previous eight-loop record set by human researchers.

This breakthrough was inspired by a challenge issued to the AI community by Matt von Hippel, a well-known science writer with a background in theoretical particle physics. Von Hippel’s challenge focused on tackling difficult problems in the field of scattering amplitudes, a key area of study in modern theoretical physics.

The Anthropic team took up this challenge and guided Claude to complete the calculations using two distinct physical pathways, employing methods such as bootstrapping. Over the course of two weeks, the results were rigorously verified by a team led by Lance Dixon, a leading expert and former world record holder in this field. The verification confirmed the accuracy of Claude’s calculations.

Almost simultaneously, a team led by He Song from the Institute of Theoretical Physics at the Chinese Academy of Sciences also completed related calculations. Using their independently developed research framework and with assistance from GPT-6, they publicly released their core data ahead of Anthropic.

While these achievements are currently based on theoretical physics “toy models” and do not yet address questions about the real universe, they represent a significant advancement. Large AI models like Claude can now deeply engage in complex deductions within cutting-edge theoretical physics, showcasing the immense potential of AI in frontier scientific research.